U.S. patent number 4,910,401 [Application Number 06/341,131] was granted by the patent office on 1990-03-20 for lwir sensor system with improved clutter rejection.
This patent grant is currently assigned to The Boeing Company. Invention is credited to Weightstill W. Woods.
United States Patent |
4,910,401 |
Woods |
March 20, 1990 |
LWIR sensor system with improved clutter rejection
Abstract
The system comprises a scanning telescope for scanning a field
of view in a vertical direction. A detector array is positioned in
the focal plane of the telescope for receiving radiant energy from
the scanned image. The detector array comprises a plurality of
elements which are positioned such that adjacent elements along the
scan direction are offset with respect to each other. A processor
receives output signals from each of the elements. The processor
delays the signals received from leading elements in the array and
adds these delayed signals to output signals from trailing elements
in the array to form pseudodetector sums. Pseudodetector sums
formed in this manner are geometrically filtered in a cross-scan
direction and also filtered in the along scan direction.
Inventors: |
Woods; Weightstill W. (Redmond,
WA) |
Assignee: |
The Boeing Company (Seattle,
WA)
|
Family
ID: |
23336358 |
Appl.
No.: |
06/341,131 |
Filed: |
January 20, 1982 |
Current U.S.
Class: |
250/332; 250/334;
348/262; 348/E3.01; 348/E5.09 |
Current CPC
Class: |
G02B
23/00 (20130101); G02B 26/12 (20130101); H04N
3/09 (20130101); H04N 5/33 (20130101) |
Current International
Class: |
G02B
23/00 (20060101); G02B 26/12 (20060101); H04N
5/33 (20060101); H04N 3/02 (20060101); H04N
3/09 (20060101); H01L 025/00 (); G02B 026/10 () |
Field of
Search: |
;250/332,334,338,342,347,349,578 ;356/1,4,141,152
;358/212,213,113,213.24,213.28 |
References Cited
[Referenced By]
U.S. Patent Documents
Primary Examiner: Buczinski; Stephen C.
Attorney, Agent or Firm: Hammar; John C.
Government Interests
The U.S. Government has rights in this invention pursuant to
contract #DASG 60-76-C-0090 awarded by the Department of the Army.
Claims
I claim:
1. An imaging system, comprising:
means for receiving energy from a field of view and for scanning
said received energy across a two-dimensional focal plane in a scan
direction;
detector means positioned in said focal plane for receiving said
energy, said detector means comprising a plurality of detector
elements, said elements being positioned in an array with elements
along said scan direction being offset transversely to the scan
direction and overlapping with respect to each other from leading
elements to trailing elements, each of said elements producing
output signals;
processing means for receiving said output signals from each of
said elements, delaying output signals from said leading elements
in said array and adding said delayed output signals to output
signals from overlapping trailing elements in said array; and
utilization means for receiving said added signals and forming
image signals in response thereto.
2. The imaging system as set forth in claim 1, wherein each of said
elements has an along scan dimension and a cross-scan dimension
which are approximately equal to the along scan dimension and the
cross-scan dimension of an optical blur formed on said focal plane
by said receiving means in response to the reception of energy from
a ar-field point source object.
3. The imaging system set forth in claim 1, wherein said elements
are disposed in a plurality of aligned rows in the cross-scan
direction.
4. The imaging system as set forth in claim 2, wherein said
adjacent elements overlap and wherein said offset is equal to the
detector-to-detector spacing in a row divided by N where N is the
number of rows in said array.
5. The imaging system as set forth in claim 4, wherein said added
signals are produced for each delayed combination of N sequentially
staggered detectors, each of said added signals being defined as
one pseudodetector sum.
6. The imaging system as set forth in claim 5 and further wherein
said processing means includes means for spatially filtering said
pseudodetector sums in the cross-scan direction.
7. The imaging system as set forth in claim 5 and further wherein
said processing means includes means for filtering said
pseudodetector sums in the scan direction.
8. The imaging system as set forth in claim 6 and further including
means for filtering said filtered sums in the said direction.
9. The imaging system as set forth in claim 1 and further including
multiplexing means for serially applying said output signals to
said processing means.
10. The imaging system as set forth in claim 9 and further wherein
said processing means includes interface adapter means for
receiving said serial output signals and storing said output
signals in predetermined memory locations.
11. The imaging system as set forth in claim 8 and further
including mean for determining peak values of said cross-scan
direction filtered and said scan direction filtered pseudodetector
sums.
12. The imaging system as set forth in claim 1, wherein said
detector means comprises a plurality of said arrays, and further
including filter means for dividing said energy into a plurality of
frequency bands, a different one of said bands being received by a
different one of said arrays, and further including a separate one
of said processing means for receiving output signals from each of
said arrays.
13. The imaging system as set forth in claim 12 and further wherein
said utilization means includes means for correlating said added
signals from each of said processing means with corresponding added
signals from other of said processing means.
14. The imaging system as set forth in claim 1, wherein said
receiving means comprises a telescope.
15. The imaging system as set forth in claim 14, wherein said
detector elements are photosensors.
16. A method for the production of signals for use in radiant
imaging of objects, comprising:
positioning a plurality of detector elements in an array with
elements in a first direction being offset transversely to the
first direction and overlapping with respect to each other from
leading elements to trailing elements;
receiving energy from a field of view;
scanning said received energy across said array in said first
direction from leading elements to trailing elements;
receiving output signals from each of said elements and delaying
output signals from said leading elements in said array and adding
said delayed output signals to output signals received from
overlapping trailing elements in said array.
17. The method as set forth in claim 16 and further including the
step of determining the dimensions of the area of said array
covered by the energy received from a distant point source target,
and forming each of said detector elements with dimensions
approximately equal to the dimensions of said area.
18. The method as set forth in claim 16, including the step of
aligning said detector elements in a plurality of rows in a
direction perpendicular to said first direction.
19. The method as set forth in claim 18 and further including the
step of making the offset of said detector elements equal to the
detector-to-detector spacing in a row divided by N where N is the
number of rows in said array.
20. The method as set forth in claim 19, wherein the step of adding
said delayed output signals to output signals from trailing
elements in said array includes the step of forming a sum for each
delayed combination of N sequentially offset detectors, each of
said sums being defined as one pseudodetector sum.
21. The method as set forth i claim 20 and further including the
step of geometrically filtering said pseudodetector sums in a
direction perpendicular to said first direction.
22. The method as set forth in claim 20 and further including the
step of filtering said pseudo-detector sums in said first
direction.
23. The method as set forth in claim 21 and further including the
step of filtering said filtered sums in said first direction.
24. The method as set forth in claim 20 and further including the
step of serially accessing each of said detector elements by a
multiplex or and reading said outputs serially
25. The method as set forth in claim 23 and further including the
step of storing said serially read output signals in predetermined
memory locations such that said pseudodetector sums can be obtained
by summing data in sequential memory locations.
Description
BACKGROUND OF THE INVENTION
This invention relates to long wave infrared radiation imaging
systems and especially to such systems which are adapted to
distinguish far-field point source objects from near-field
background clutter.
Presently, a refracting or reflecting telescope is used for the
sensing, isolation and the characterization of radiation from
far-field "point" sources. The telescope is boosted aloft in a
rocket, aircraft or similar vehicle. The vehicle is positioned to
view a threat train which may contain incoming targets such as
hostile nuclear missiles. These targets are first detected at a
great distance and appear as point source images to the telescope.
The field of view of the telescope is imaged onto a focal surface
and scanned across that surface with the intent of observing one or
more such targets in a track-while-scan mode.
The concept is applicable to both electromagnetic and acoustic
radiation. Long wave infrared wave lengths are preferentially
employed in this task, as radiation detection devices in this
region are available and capable of detection of a small number of
photons per second emitted by a warm object at distances of several
hundred miles. However, the combined effects of thermal radiation
and the reflection of sunlight and earthshine and albedo produce
distinguishable patterns for objects or clouds of objects, as
characterized by angular extent, magnitude, and wave length
distribution or "color". A typical detector configuration located
at the focal plane of the telescope comprises individual detector
elements consisting of separate semiconductor chips mounted to
circuit boards upon which the wiring, load resistors and low-noise
electronic amplifier elements are also mounted. The detectors are
grouped typically in a series of two to four arrays, each with its
own optical bandpass filter to allow characterization of incoming
radiation according to wavelength distribution. Images of far-field
objects are scanned across the arrays, either by mechanical motion
of the telescope or by rotation of one or more optical elements
within the telescope.
The image size of a far-field point source object is, in general,
limited by diffraction of the telescope optical system, which
produces a characteristic blur. The size of the detector elements
has been determined based upon the size of the characteristic blur.
Conventionally, the along-scan dimension of the detector elements
is sized to match the along scan dimension of the blur. There are
two reasons for this size consideration. Detector elements of
significantly greater dimension than the blur produce signals of
much poorer spatial resolution so that the position uncertainty of
objects increases. When the detector elements are made small with
respect to the blur, less radiation is intercepted by the detector,
and the signal levels increase. Inasmuch as the noise floor for
signal detection is generally limited by the amplifier and not the
detector, the use of smaller detectors degrades the minimum size of
objects which can be sensed and the range at which the objects can
be detected.
In the cross-scan direction, other criteria must be considered. In
order to sense and characterize all objects in a threat train, a
significant field of view must be encompassed. The detector
elements must in combination sweep out this field of view. On the
other hand, the number of detectors is limited by the number of
electronic channels which can be crowded into the focal plane array
using available technology. A further limitation is imposed by the
data processing hardware which can perform the needed pulse
extraction, correlation and other data processing in real-time.
Accordingly, to fill the field of view in several wavelength bands,
the detectors have traditionally been made long in the cross-scan
direction. These considerations lead to a detector length several
times the minimum resolution element. In order to allow the
determination of cross-scan position to an uncertainty small with
respect to detector length, the positions of the detectors are
staggered in the along scan direction.
Several difficulties are inherent in these detectors. First, when
two targets approach each other at the same elevation position, the
signals add together on a given detector and it becomes difficult
to separate their radiometric qualities and pursue their magnitude
and color classification. Second, since the detectors are several
times the size of the point source image blur, their response to
defocused clutter signals is relatively large. Finally, when an
image crosses the end of a detector, the proportion of image energy
falling on or off the detector is difficult to determine, and such
measurements are generally discarded in radiometric value
determination.
Improvement of the state of art of focal plane detector fabrication
and signal processing hardware has lead to studies, proposals, and
research into focal plane assemblies with large numbers of
detectors, up to the thousands or hundreds of thousands. Typically,
however, these arrays, or mosaics as they are called, are laid out
in orthogonal rows and columns. To overcome the problems with
partial image coverage, the detector sizes are made small with
respect to the image blur size, leading, with reasonable fields of
view, to detector numbers in the millions. Present readout and data
processing technology is not capable of handling this
arrangement.
Accordingly, a need has arisen for a focal plane detector array and
a processing scheme which can overcome the above-discussed
deficiencies and yet is capable of operation with present
processing technology.
SUMMARY OF THE INVENTION
A principal object of the present invention is to provide a radiant
imaging system which is capable of distinguishing point source
targets from background, illumination and near-field particle
clutter.
A further object of the present invention is to provide a radiant
imaging system utilizing focal plane detector arrays wherein each
detector is sized so as to produce a high degree of resolution and
rejection of near-field clutter and yet wherein the array
information output is compatible with present data processing
techniques.
Another object of the present invention is to provide a radiant
imaging system utilizing focal plane detector arrays wherein the
detectors are oriented with respect to one another in a manner
which allows maximum utilization of the information received
therefrom.
Yet another object of the present invention is to provide a radiant
imaging system which includes a processing technique wherein the
time delayed summation of outputs of detectors in the focal plane
detector array is made to form an equivalent mathematical single
row of pseudodetectors of closer spacing, enabling the enhancement
of signal-to-noise ratio.
A further object of the present invention is to provide a radiant
imaging system in which data from a series of detector of the focal
plane detector array is processed similarly in two orthogonal
dimensions, the first dimension being purely geometrical in time
stasis such that signals from point sources are extracted
unambiguously from rapidly scintillating focused sources and
background signals.
Another object of the present invention is to provide a radiant
imaging system which has increased sensitivity for distinguishing
between signals from two closely spaced objects.
In accordance with these and other objects, the radiant imaging
system of the present invention comprises a means for optically
scanning a field of view along a scan direction and applying
received radiant energy onto a two dimensional focal plane.
Detector means positioned in the focal plane receives the radiant
energy. The detector means comprises a plurality of detector
elements which are positioned in an array with adjacent elements
along the scan direction being offset with respect to each other.
Each of the elements produces an output signal which is delivered
to a processing means. The processing means delays output signals
from leading elements in the array and adds the delayed output
signals to output signals from trailing elements in the array.
In accordance with a further aspect of the invention, the system is
adapted to receive radiant energy from distant objects and to form
diffraction-limited optical blurs on the focal plane. Each of the
elements of the array has an along scan dimension approximately
equal to the along scan dimension of the diffraction-limited
optical blur and has a cross-scan dimension approximately equal to
the cross-scan dimension of the diffraction-limited optical
blur.
The elements are disposed in a plurality of aligned rows in the
cross-scan direction. The adjacent elements in the along scan
direction overlap and are offset by an amount equal to the detector
spacing in a row divided by n, where n is the number of rows in the
array.
The processing means includes means for forming a sum for each
delayed combination of n sequentially staggered detectors. Each of
the sums is defined as one pseudodetector sum. The processing means
further includes means for geometrically filtering the
pseudodetector sums in the cross-scan direction and also includes
means for filtering the pseudodetector sums in the along scan
direction.
BRIEF DESCRIPTION OF THE DRAWINGS
The above and other objects will become more readily apparent as
the invention becomes more fully understood based on the following
detailed description taken in connection with the drawings in which
like numerals represent like parts throughout and in which:
FIG. 1 is a schematic view of an optical scanner which may be used
in the present invention;
FIG. 2 is a schematic view of three detector arrays to be used in
the present invention for detecting the three primary colors;
FIG. 3 is a schematic representation showing an optical blur
representing a far-field point source object;
FIG. 4 is a schematic representation of one detector of a focal
plane detector array;
FIG. 5 is a schematic representation showing the orientation of
detectors in one focal plane detector array;
FIG. 6 is a chart indicating the formation of pseudodetector sums
from the detectors of FIG. 5;
FIG. 7 is a block diagram depicting the operation of the system of
the present invention;
FIG. 8 is a schematic representation showing a field of view image
crossing the first row of the detector array of FIG. 5;
FIG. 9 is a schematic representation showing a field of view image
crossing the second row of detectors of the detector array of FIG.
5;
FIG. 10 is a schematic representation showing a field of view image
crossing the third row of detectors of the detector array of FIG.
5;
FIG. 11 is a schematic representation showing a field of view image
crossing the fourth row of detectors of the detector array of FIG.
5;
FIG. 12 is a flow diagram showing the steps to be carried out in
the interface adapter of FIG. 7;
FIG. 13 is a data storage map showing the preferred manner of data
storage produced by the interface adapter;
FIG. 14 is an updated data storage map similar to that of FIG.
13;
FIG. 15 is an updated data storage map similar to that of FIG.
14;
FIG. 16 is a flow diagram showing the steps to be carried out in
the pseudodetector sum block of FIG. 7;
FIG. 17 is a flow diagram showing the steps to be carried out in
the cross-scan filter block of FIG. 7;
FIGS. 18-A and 18-B show a flow diagram depicting the steps to be
carried out in the along scan filter block of FIG. 7; and
FIGS. 19-A and 19-B show a flow diagram depicting the steps to be
carried out in the two-dimensional peak find block of FIG. 7.
DESCRIPTION OF THE PREFERRED EMBODIMENT
FIG. 1 shows a scanning apparatus for scanning received radiant
energy shown as 111 across a plurality of photodetector arrays 131,
132 and 133. The apparatus itself is a standard device which forms
no part of the present invention and is shown for exemplary
purposes only.
Radiant energy 111 is received in a telescope or the like and
reflected by mirror 116 onto rotating multifaceted mirror 112. As
mirror 112 rotates, the received energy is reflected across
photodetector arrays 131, 132 and 133 which are placed in the focal
plane of the receiving telescope. Optical filters 119, 120 and 121
are placed between mirror 112 and detector arrays 131, 132 and 133,
respectively. Filters 119, 120 and 121 are spectral band filters
such that the images generated by detectors 131, 132 and 133
represent the source image energy in the short, medium, and long
wavelengths, respectively.
FIG. 2 shows a more detailed view of the detector arrays 131, 132
and 133. Each array consists of a plurality of rows of detectors.
The number of rows to be used and the number of detectors in each
row are determined in accordance with design constraints to be
discussed hereinafter. It should be noted that each array is
identical to the other arrays. The only difference between arrays
is in that a different wavelength band of radiant energy is
received by each. For example, array 131 may receive only short
wavelength radiation while array 132 receives only medium
wavelength radiation and array 133 receives only long wavelength
radiation.
With this in mind, it will be understood that the present invention
will be described with reference to only array 131 with the
understanding that the configuration of the other arrays and
processing techniques to be used therewith are identical to those
discussed with respect to array 131.
FIG. 5 shows array 131 in greater detail. Preferably, array 131
contains several hundreds or thousands of individual detectors, 40
of which are shown as detectors D-1 through D-40. The detectors of
the array are divided into four or more rows. The scanning
apparatus of FIG. 1 causes a received field of view image to be
scanned vertically along the arrays in he along scan direction as
depicted by arrow 135 shown in FIGS. 2 and 5. The rows of detectors
are aligned in the cross-scan direction defined as the direction
perpendicular to arrow 35. In the along scan direction, the rows
are offset by an amount equal to the detector-to-detector spacing
within a row divided by the number of rows in the array.
As discussed above, the system of the present invention is designed
to image far-field targets whose angular substance is small
compared to the diffraction limit of the sensor telescope optics.
Under these conditions, the targets are described as point sources.
The number of detectors and the size of the detectors must be
determined in relation to the particular image to be received. Due
to optical effects of the telescope being used, a point source
distant target image appears on the focal plane of the telescope as
a characteristic blur. As shown in FIG. 3, the characteristic blur
may have a general shape in which the along scan dimension B of the
blur is approximately twice the cross-scan dimension A. Each
detector of the detector array, as typified by the single detector
shown in FIG. 4, has an along scan dimension approximately equal to
the along scan dimension B of the characteristic blur and a
cross-scan dimension which is approximately equal to the cross-scan
dimension A of the characteristic blur. Accordingly, in this
example the along scan dimension of each detector is approximately
twice the cross-scan dimension of the detector.
Each detector can be a charge coupled device or other standard
photosensitive element. The output of each detector is amplified or
conditioned prior to processing. As mentioned above, the noise
floor for signal detection is limited by the amplifier and
generally not the detector. Accordingly, a reduction in size of the
detectors in the along scan dimension below that of the blur along
scan dimension would degrade the minimum size of objects which
could be sensed and the range at which they could be detected since
the noise level of the amplifier would not be reduced with the
reduction in along scan dimension of the detector. An increase in
size of the detector in the along scan dimension to a significantly
greater dimension than the blur would produce signals of poor
spatial resolution so that the position uncertainty of objects
would increase. Due to these considerations, the along scan
dimension of the detectors is made roughly equal to that of the
envelope enclosing 80% of the energy of the blur.
The cross-scan dimension of the detectors is dictated b the field
of view to be encompassed balanced by the number of channels which
can be crowded into the focal plane area together with processing
limitations of the data processing hardware which is used. The
present invention uses detectors which have a cross-scan dimension
equal to the blur cross-scan dimension in order to balance these
factors and to provide a low ratio of clutter-to-point source image
signals. Also, the detector size enables a powerful method for
extracting the point source image signals from the reduced clutter
signals to be used, as will be discussed.
In order to adequately fill the field of view, a large number of
detectors is included in each row of the detector array. For the
example to be discussed, each row will contain 108 detectors. The
spacing between detectors is made to be approximately 0.2 blur
widths in order to provide adequate detector separation. Therefore,
the detectors in a given row are 1.2 blur widths spacing center to
center.
The number of rows in the array determines the number of cross-scan
samples of data to be processed. Accordingly, as the number of rows
per array increases, the uncertainty in peak determination
decreases. Two rows per array appeared to yield excessive
uncertainty in peak determination and thus four rows per array has
been chosen to be the minimum preferred. With four cross-scan
samples per blur, the ideal peak magnitude uncertainty is reduced
to less than 2 percent.
For purposes of further explanation, it will be assumed that each
detector array has four rows of detectors. It will be understood
that this type of focal plane design and processing design may be
extended to five, eight or more rows of detectors with commensurate
reduction of noise equivalent flux density and processing
uncertainty as well as increased azimuthal resolution and data
processing load. At larger row numbers, the field of view may
become over-filled by the focal plan and data rates may become
excessive.
As shown in FIG. 5, the row-to-row offset is 0.3 blur widths or an
overlap of 0.7 blur widths. Offsetting the rows allows a
determination of cross-scan position uncertainty which is
small.
Time delay integration (TDI) of a multi-row array of detectors is
often employed to enhance signal-to-noise ratio or reduce noise
equivalent flux density in arrays having no row to row offset. The
concept of TDI processing is applied to the staggered row array of
FIG. 5 by forming pseudodetector sums which are set forth in the
chart of FIG. 6. As is apparent from FIG. 6, one pseudodetector sum
is formed for each group of sequentially offset detector elements.
The pseudodetector can be visualized as having a theoretical width
equal to the width of a detector divided by the number of rows in
an array and aligned in rows in the same manner as the array. In
order to more fully understand the TDI concept as applied to array
131 reference is had to FIGS. 8-11. In FIG. 8, it will be seen that
the field of view image 164 is beginning to be scanned over the
detector array. The first portion of image 164 overlies the first
row of detectors. At this time, a first time sample of the outputs
of all detectors is taken. The data received is referred to as
frame 1. FIG. 8 shows the image being scanned across the second row
of detectors, row 166. Frame 2 information is obtained at this
point in time. FIG. 9 shows the image being scanned across the
third row, row 167. Frame 3 information is obtained at this point
in time. FIG. 10 shows frame 4 information being obtained as the
image 164 is scanned across row 168. From FIG. 6, it will be
understood that pseudodetector sum P1 is formed by adding the
outputs of detector D1 during frame 1, D2 during frame 2, D3 during
frame 3, and D4 during frame 4. Each pseudodetector sum is formed
in a similar manner to represent the summation of the outputs from
our adjacent offset detectors taken from similar portions of image
164 during the scan of image 164. It should be understood that this
description relates only to the first row of pseudodetector sums
and that additional rows of pseudodetector sums are formed in a
similar manner for additional portions of the image. The total
number of pseudodetector sums in each row is equal to the total
number of detectors in the array less N-1, where N is the number of
detector rows. In the specific example set forth, 429
pseudo-detector sums are formed for each array.
FIG. 7 sets forth a block diagram describing the complete operation
of the system of the present invention. A radiating far-field
object 138 emits radiant energy which is picked up by the sensor
telescope optics 140. Scanning system 110, one possible
implementation of which is show in greater detail in FIG. 1, causes
the sensed image to be scanned along the focal plane detector array
131 which is shown in greater detail in FIG. 5. Each frame of
information is read out serially from the detectors of array 131 by
read-out multiplexer 142. The frequency at which read-out
multiplexer 142 reads frames of information from array 131
determines the maximum product of resolution and scan rate
obtainable.
The time that it takes for one portion of image 164 to traverse one
row of detectors, for example, the time it takes for the leading
edge of image 164 to traverse row 165 as shown in FIG. 8, is
referred to as a dwell. For adequate image resolution, a minimum of
four samples per dwell to produce four frames of information per
dwell is required. In the example discussed above, only one sample
per dwell was described for simplicity.
The outputs from array 131 are in analog form. These outputs are
received serially by analog-to-digital converter 144 and converted
to digital form. An interface adapter 145 receives the digital
output signals and assigns them to specific storage locations in
digital RAM 148.
The data in RAM 148 is received by the microprocessor of block 150
which forms the pseudo-detector sums by TDI as shown in FIG. 6. The
pseudodetector sums are stored and geometrically cross-scan
filtered using a digital matched filter routine which can be
implemented by a second order sine x/x bandpass filter.
In regard to cross-scan filtering of the pseudodetector sums, it
should be clearly understood that each row of 429 pseudodetector
sums is filtered individually. For example, referring again to
FIGS. 8-11 wherein the imaging process using one sample per dwell
is described for simplicity, it can be seen that image 164 is
divided into four image sections referred to as 164a, 164b, 164c
and 144d. The formation of pseudodetector sums described above was
in reference to the pseudodetector sums for image section 164a. In
other words, after image section 164a has traversed array 131 from
row 165 to row 168, the first row of pseudodetector sums is formed.
Similarly, a second row of pseudodetector sums is formed after
image section 164b has completely traversed array 131 from row 165
through row 168. A third row of pseudodetector sums is formed for
image section 164c and a fourth row is formed for image section
164d. Clearly, since the minimum number of samples per dwell to be
used for adequate image resolution is four, a total of 16 rows of
pseudo-detector sums must be formed for 16 image sections. In this
case, the image sections necessarily overlap. Each row of
pseudodetector sums is filtered using a symmetrical, linear phase,
bandpass filter approximating a matched filter to the geometric
shape of the optical blur pattern in the cross-scan direction. This
processing detects the position and magnitude of point source
signals while suppressing signals from the broad blurs produced by
near-field objects and other non-point source phenomenon. The
specific filter to be used can be a four sample by eight sample,
second order sine x/x digital bandpass filter.
Again with reference to FIG. 7, it will be seen that after all rows
of pseudodetector sums are cross scan filtered, the pseudodetector
sums are along scan filtered. The along scan filtering is performed
on the cross-scan-filtered pseudodetector sums which are produced
at successive time samples. Accordingly, it will be understood that
a total of 429 along scan filter routines must be performed
simultaneously. These routines are performed in parallel using one
or more microprocessors. The along scan filters used are similar to
the cross-scan filters used and can be implemented by second order
digital sine x/x bandpass filters.
Since sources in the near-field typically vary rapidly in time, the
scintillation effects caused by the sources are frozen with each
snapshot which is in effect taken by storing outputs from each of
the detectors. Accordingly, in order to extract signals from point
sources unambiguously from the rapidly scintillating near-field
broad sources and background signals, cross-scan filtering is
performed first since this dimension is purely geometrical in time
stasis. If along scan filtering is performed before cross-scan
filtering, the point source images may not be extracted as
accurately.
After the cross-scan and along scan filtering routines, a peak find
routine is performed at block 154. The peak find routine determines
the maximum between two successive threshold crossings of the
filtered pseudodetectorsums. The peak find routine includes
identical routines which are performed in both the along scan and
cross-scan directions. In each routine, the peak values are
compared to a non-zero threshold and only those exceeding the
threshold are retained. Ideally, with no noise, there would be four
adjacent equal peak values in the cross-scan direction all with the
same along scan address. Four peak values appear due to the fact
that each pseudo-detector sum theoretically has a cross-scan
dimension which is one-quarter of the width of one characteristic
blur when four rows are used. When noise is added to the detector
signals, the peaks will scatter in amplitude and along scan
location. The simplest method for choosing the correct peak value
in address is to describe a box of a predetermined number of
locations about a non-zero peak. All but the maximum value within
the box are eliminated and the box is moved to the maximum value
and the process repeated. If two or more equal values appear within
the box, they are assumed to be from the same target. That is, the
value is taken for the peak and the centroid of their locations is
taken for the address.
Object data from other color bands is determined in a similar
manner indicated at box 158. Array to array correlation is
performed at 156. In this routine, an object must appear in two or
more arrays or be discarded.
The information obtained at block 156 for each scan is stored and
correlated with previous scans in correlator 160. The object
history produced by correlator 160 is fed to discrimination and
tracking circuit 162. The object's trajectory can therefore be
predicted and the object intercepted.
FIG. 12 shows a flow diagram for one form of an interface adapter
146. The routine of FIG. 12 should be performed for each scan. At
step 200, the routine is entered. At step 201, the scan direction
is determined. For the example set forth, the scan direction never
changes nd is always vertically upward. However, the routine is
designed to operate for both up and downward movement in the
vertical scan direction. At step 202, the pointer increment is
obtained. The pointer increment is determined so that the data
received by the interface adapter is organized according to the
preferred delays for a four-row array with four samples per dwell,
as will be discussed below. Step 203 initializes the frame number
and pointer. Step 204 determines whether a new data frame has been
started or not. If a new data frame has not been started, the
program waits for the beginning of the next data frame. Step 205
increments the frame number. Step 206 calculates the pointer which
determines the memory location in which the incoming data is to be
stored. Step 207 determines whether data is available to be stored.
If not, the program waits for data to be supplied. Step 208 stores
the data in the calculated memory location. Step 209 determines if
all data from the present frame has been stored. If not, the
program returns to step 206 an recalculates the pointer to store
the next incoming data. If, at step 209, the end of a frame has
been reached, the program proceeds to step 210 which determines the
scan direction. If the scan direction has changed, the program must
return to step 202 in order to obtain the new pointer increment. If
the scan direction has to changed, the program proceeds to step 205
to increment the frame number and continues to store data in the
next succeeding frame.
FIG. 13 shows a data storage map as it would appear while data from
the 14th frame is being read into the memory locations by the
interface adapter of FIG. 12. The memory locations are indicated in
octal numbering. The vertical axis of the data storage map
indicates the memory start locations. The horizontal axis indicates
the specific memory locations. In parentheses below each memory
location the detector element from which the data was taken is
indicated. Memory locations are only included for 16 detectors, it
being understood that there would be sufficient memory locations to
store data from each detector of the array. In the present example,
there are 432 detectors. The data storage map of FIG. 13 is
calculated based upon there being four time samples per dwell to
produce four frames of information per dwell. Accordingly, 16
frames of information are obtained in the time that it tasks for
the field of view image to be scanned across all four rows of the
array. Therefore, 16 memory locations are allocated to each
detector. The interface adapter calculates the pointer such that
data from detector D1 during the first frame is stored in memory
location 000. As seen in FIG. 13, data from the second frame is
stored in location 200 (OCTAL). Data from successive frames fill
memory locations vertically on the map. Data from detector D2 from
the first frame is stored in location 3001 (OCTAL) and data from
successive frames fill vertically adjacent memory locations up to
location 3601 (OCTAL). Data from the fifth frame is stored in
location 001 (OCTAL). Data for each detector is similarly offset
such that each row of memory locations stores data for each set of
four sequentially numbered detectors, which data is offset by four
time samples. Accordingly, with reference to FIG. 5, it can be seen
that data from each row of detectors, for example row 165, taken
during the same time sample, is stored in the same row and data
from adjacent rows taken four time samples later is stored in the
same row. Consequently, the data in each memory location row
corresponds to data frames taken from the same image section of the
field of view image as it is scanned across each row of detectors.
The data from each row can be read out sequentially and added in a
trivial manner to produce the pseudodetector sums.
FIG. 13 depicts the situation where data from detector D5 at the
14th data frame is about to be entered in location 3200 (OCTAL).
The data from the 13th frame has all been entered and the first row
of memory locations are ready to be read out by microprocessor of
block 150 to for the first pseudo-detector sums. Similarly, FIG. 14
depicts the storage map after the 15th data frame has been
completely read in and data during the 16th storage frame for
detector D5 is being read into memory location 3604 (OCTAL). At
this point in time, the third row of memory locations is being read
out by microprocessor 150 to form the third set of pseudodetector
sums. In a similar manner, FIG. 15 shows the data storage map
during storage of data from frame 18. The data for that frame from
detector D5 is about to be overwritten into location 204 (OCTAL).
At this point in time, the first portion of tee field of view image
has been scanned across all the rows of detectors and all
pseudo-detectors associated with the first portion of the image
have been calculated by the microprocessor. Accordingly, the
storage locations associated therewith, comprising the first four
rows of memory locations, can be overwritten.
Clearly, it can be seen that the interface adapter organizes the
data according to the preferred delays for a four-row array with
four time samples per dwell. The rearrangement for larger numbers
of rows or greater time sampling rates should be obvious. It should
be noted that since the memory locations are overwritten in stages,
only a few storage rows are valid at any one time. Accordingly,
microprocessor 115 must keep up with the interface adapter or it
will get bad data. The alternative to this storage arrangement
would be to store each frame of data in sequential memory cells and
have the microprocessor calculate each pointer to access the
properly delayed data. As the processor is quite busy, however, it
usually makes more sense to unload this task on the interface
adapter.
FIG. 16 shows a flow diagram for the time delay integration routine
performed by microprocessor 150 for forming the pseudodetector
sums. Step 300 of the flow diagram is the point at which the
routine is entered. The routine is performed 429 times to produce
429 pseudodetector sums for each row of the time delay storage map
of FIG. 13. The routine is first performed for each scan after the
1th data frame has been entered into the storage map. At step 302,
the routine determines whether the data frame is ready. If the data
frame is not ready, the routine waits until all entries int the
storage map have been made for the data frame of interest. Step 304
obtains the data frame number, initializes the pointers and
initializes the accumulator. Initially, the sum of the data for
detectors D1, D2, D3 and D4 is stored in file Y (l,j) and the
pointer is set at the location for detector D5 in the row of
interest. Step 306 calculates a pseudo-detector sum. Essentially,
the pseudodetector summation takes the form of a running average
represented by:
where n is the number of rows in the detector array, Y is the
pseudodetector sum, X is the detector data, i is the detector
number and j is the frame number. Accordingly, it will be
understood that each pseudo-detector sum Y is the summation of the
outputs of four sequentially numbered detectors formed by adding
the data in n sequential memory locations of a row indicated in the
storage maps of FIGS. 13, 14 and 15. Step 308 stores each
sequentially formed pseudo-detector sum in a rotating frame file.
Step 310 increments the pointers. Step 312 determines whether the
end of the frame has been reached. If the end of the frame has not
been reached, the routine returns to step 306 where the next
pseudodetector sum is formed. If the end of the frame has been
reached, the routine returns to step 302 to perform the summation
process on the row of memory locations in FIG. 13 completed by the
next data frame.
FIG. 17 shows the cross-scan filter routine to be performed on the
pseudodetector sums formed by the routine of FIG. 16. At step 500,
the cross-scan filter routine is initiated. At step 502, the
accumulators are initialized. At step 504, the routine determines
if all time delay integration steps have been performed and thus
all pseudo-detector sums have been calculated for a particular
frame of information. A band-pass filter is implemented by taking
the difference between two low-pass filters with a ratio of 2:1 in
cut-off frequencies. The low pass filters used in the example are
integer coefficient, recursive sine x/x filters.
The first low-pass filter output is determined by the mathematical
equations:
where AA is the band-pass filer output and A is an intermediate
quantity determined to facilitate calculation.
The second low-pass filter has a pass-band larger than AA and is
determined by the equations:
where BB is the filter output and B is an intermediate numerical
step provided for ease of calculation.
The optimum value f m for the best approximation to a matched
filter for the target pulse should be approximately 1/3 the number
of samples per dwell. In the example chosen, with approximately 4
samples per dwell in the cross scan direction, m=1.
The band-pass filter output is given by the mathematical
formula:
where C is the band-pass filter output. Note that inverse scaling
factors (4) and (4 m.sup.2) must be applied to retain original
dimensions.
In equations 2-6, Y indicates a pseudo-detector sum, i indicates
detector or pseudodetector number, j indicates from number and m
indicates the filter width parameter which determines the response
of the filter.
Step 508 stores each bandpass filtered pseudo-detector sum C. Step
510 determines whether the end of the frame has been reached. If
not, step 512 increments the detector number i add returns the
routine to step 506. If the end of the frame has been reached, the
frame number j is incremented and the detector number i is returned
to 1. The routine is then retuned to step 502 to filter the next
row of pseudo-detector sums.
FIGS. 18A and 18B show an along scan filter and pulse shape
discriminating routine. Step 601 initiate the routine which is
carried out for each filtered pseudodetector sum generated by the
routine of FIG. 17. Step 602 initializes the accumulators, buffers
and pointers to be used in the routine. Step 604 initializes the
detector pointer i to 1. Step 606 determines whether all
information for the data frame of interest has been generated. If
not, the routine waits for the data frame information to be
generated. If the data frame is ready, control is passed to step
608 which comprises two digital bandpass filters formed from three
low-pass filter algorithms. The three low-pass filter algorithms
have cut-off frequencies with the ratio 4:2:1.
The filter with the lowest cut-off frequency is defined by the
following equations:
where DD indicates the first low-pass filter output and D
represents an intermediate value generated to facilitate
calculation.
The low-pass filter having an intermediate cut-off frequency is
determined as follows:
where EE represents the second low-pass filter output and E
represents an intermediate result generated to facilitate
calculation.
The low-pass filter having the highest cutoff frequency is
determined by the equations:
where FF is the output of the third low-pass filter and F is an
intermediate result generated to facilitate calculation.
The first band-pass filter is provided by taking the difference
between the two low-pass filters with a ratio of 2:1 in cut-off
frequencies, as follows:
where G is the output of the band-pass filter.
The second band-pass filter is formed by taking the difference of
the low-pass filters having cut-off frequency ratios 4:2 as
follows:
where H is the output of the band-pass filter.
In equations 7-14, i,j and m represent the same parameters as
discussed with respect to equations 2-6.
Pulse width may be characterized by the ratios of the outputs G and
H of the two band-pass filters at the peak of a pulse. This ratio
characterizes the spatial frequency distribution of the target
pulse, and may be used as a discriminant for distinguishing between
targets. If desired, this discrimination method may be applied
similarly to the cross-scan filtering process. Pulse width gates of
plus or minus 40 percent of nominal may be implemented by a simple
compare, shift and compare sequence as set forth in FIG. 18B.
Peak finding and thresholding proceed as follows: Step 610 compares
G with zero, if G is greater than zero, step 624 compares the value
for G with the previous recorded maximum for G. If G is less than
or equal to the previous recorded maximum, the routine skips to
step 628 which determines the end of the frame. If the end of the
frame has been reached, the frame number j is incremented by one
and detector pointer i is reset to one at step 629 after which
control is returned to step 606. If the end of the frame has not
been reached, the pointer i is incremented by one and control is
returned of step 608.
If, at step 624, it was determined that G is greater than the
previous value for the maximum of G, step 626 records G as the new
maximum value, records the frame number j in register T, and
records the value of H as the new maximum for H. Control is then
passed to step 628.
If, at step 610, t was determined that the value of G is less than
or equal to zero, step 612 determines whether the recorded value
for the maximum of G is greater than or equal to zero. If it is
equal to zero, the routine determines that no pulse has occurred
and control is passed to step 628. If the maximum value of G is
greater than zero, control is passed to step 614 which compares the
maximum value with a predetermined threshold. If the maximum is
less than the threshold, the routine determines that the pulse peak
is too low to indicate a point source object and control is passed
to step 628. If the maximum value of G is greater than or equal to
the threshold, step 616 compares the maximum value of G with the
maximum value of H. If the maximum of G is less than the maximum of
H, it is determined that the pulse is too narrow to be a point
source object and control is passed to step 628. If the maximum of
G is greater than or equal to the maximum of H, control is passed
to step 618 which compares the maximum of G with two times the
maximum of H. If the maximum of G is determined to be greater, the
routine determines that the pulse is too wide to be a point source
object and control is passed to step 628. Otherwise, control is
passed to step 620 which permanently records the values of the
maximum of G, T and i. Step 622 then resets the register storing
the maximum of G value to zero and passes control to step 628.
As can be seen from the above description, in this implementation,
the time delay integration formation of the pseudodetector sums is
separated from the cross-scan and along scan filters, and the
individual filter components are also separated. Depending on the
hardware architecture, it may be faster to combine the
pseudodetector sum and cross-scan filter algorithms into one
combined algorithm form. Also, it should be noted that the filter
algorithm itself is not unique to the process and other filters may
be employed.
A plurality of along scan filter routines of FIGS. 18A and 18B are
performed in parallel at block 152 of FIG. 7. For the example under
discussion, 429 such parallel routines are performed and the
results are passed to peak find block 154. FIGS. 19A and 19B show a
flow diagram of a cross-scan peak find routine o be carried out at
block 154 of FIG. 7.
In FIG. 19A, step 700 initiates the routine. The routine is carried
out for each filtered data frame of pseudodetector sums. Step 702
initializes an object file A. Step 704 determines whether the
filtered data frame is ready. If the data frame is not ready, the
routine waits for all filtering operations on the data frame to be
carried out. Step 706 reads in the next maximum G value as
determined by the along scan filter routine of FIGS. 18A and 18B.
That value is assigned to register B. The frame number j and the
detector pointer i are assigned to registers j.sub.b and i.sub.b,
respectively. Step 708 compares the cross-scan distance between the
location of the presently read maximum value of G, stored at
i.sub.B and the location of the previous maximum value of G, stored
at i.sub.A. This distance is compared to a minimum space. If the
distance is less than a minimum space, step 710 determines the
relative values of B and A. If B is greater than or equal to A, the
object file is updated by making B the new value of A and making
i.sub.b the new value for i.sub.a. The routine then returns to step
706 to continue the comparison with additional values of G. If B is
less than A, the object file is not updated and the routine returns
to step 706.
If, at step 708, the distance between i.sub.B and i.sub.A is
greater than or equal to the minimum space defined, this would
indicate that B is a separate object. However, background and noise
or asymmetry of a resolved target will cause peaks from a given
target to appear in different data frames. This tends to produce
"shadow" objects which must be eliminated. To eliminate them, all
peaks within a given solid angle are compared, and the largest
value is selected and recorded as an object along with its
coordinates by the routine of FIG. 19B. In FIG. 19B, step 714 sets
a pointer to the first object OB located in the scan. Step 716
compares OB to zero. If OB equals zero, OB is updated along with
its coordinates in step 724 to be equal to B and its coordinates.
If OB is greater than zero, step 718 compares the cross-scan
distance between B and OB with a predetermined minimum space. If
the distance is greater than or equal to the minimum space, the
routine goes to step 719 which increments the pointer to the next
object and returns to step 716 to compare the next object with
zero. If the distance calculated in step 718 is less than the
minimum space, the routine goes to step 720 which calculates the
along scan distance between B and OB. If this distance is greater
than or equal to the minimum defined space, the outline goes to
step 719. If this distance is less than the minimum space, the
routine advances to step 722 to determine which of B and OB is the
actual object. A direct comparison between B and OB is made. If OB
is greater, the routine advances to step 724 where OB is updated
with the value of B. From step 724, the routine advances to step
712 where A and i.sub.A are updated with the values of B and
i.sub.B. If step 722 determines that B is less than OB, B is
determined to be a shadow and its value is not stored as an object.
The routine advances to step 712 where the value of A and its
address are updated with B and its address. The routine then
returns to step 706.
Referring again to FIG. 7, array-to-array correlator 156 receives
one magnitude and two orthogonal addresses for each point source in
each of three arrays developed by peak fin routines 154 for each
array. Although not shown, a process somewhat similar to that
carried out in the peak find routine of FIGS. 19A and 19B is
followed for array-to-array correlation except that generally an
object must appear in two or more arrays or be discarded. This
routine would be obvious to one of ordinary skill in the art.
Similarly, the scan-to-scan correlator 160 and discrimination and
tracking routine 162 would be obvious to one of ordinary skill in
the art and will not be disclosed in detail here.
The above description is set forth for purposes of illustration
only. As should be apparent, numerous modifications, additions and
variations of the present invention may be made without departing
from the scope thereof as defined by the appended claims.
* * * * *